"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629964487.380869/events.out.tfevents.1629964487.8e89bd551565.924.111 +3 -0
- model-bin/finetune/base/log/1629964926.187713/events.out.tfevents.1629964926.8e89bd551565.924.113 +3 -0
- model-bin/finetune/base/log/1629965370.4697628/events.out.tfevents.1629965370.8e89bd551565.924.115 +3 -0
- model-bin/finetune/base/log/1629965797.9083533/events.out.tfevents.1629965797.8e89bd551565.924.117 +3 -0
- model-bin/finetune/base/log/1629966225.1204813/events.out.tfevents.1629966225.8e89bd551565.924.119 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629964487.8e89bd551565.924.110 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629964926.8e89bd551565.924.112 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629965368.8e89bd551565.924.114 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629965797.8e89bd551565.924.116 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629966225.8e89bd551565.924.118 +3 -0
model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165393
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0f10fcf2ba5b5043370eb75b02ab550fa8de52a8d48b785cd2a7a34ee8cd3764
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/pytorch_model.bin
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 377909911
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:03328ac6ccc2b112de19e4fb255a6b080339e8765bde06155a660382a04b6fd0
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14503
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:836e20a6bf2106602ba0bcf5f0a522df38ed03fb672c5d7ac728e418b994c5c5
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/scaler.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 559
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1489a51728ff137e38a35bf5e8ede61ad98416177be49f5b464465239ddbc43f
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/scheduler.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 623
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:72dea23aa8c63e88416a4f9fa0ccbab7dee5eb283cd5168ed9449f3c30f52e3b
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -220491,11 +220491,800 @@
|
|
| 220491 |
"eval_steps_per_second": 0.695,
|
| 220492 |
"eval_wer": 0.18689181453921008,
|
| 220493 |
"step": 95203
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 220494 |
}
|
| 220495 |
],
|
| 220496 |
"max_steps": 620000,
|
| 220497 |
"num_train_epochs": 5000,
|
| 220498 |
-
"total_flos": 2.
|
| 220499 |
"trial_name": null,
|
| 220500 |
"trial_params": null
|
| 220501 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
+
"epoch": 771.995983935743,
|
| 5 |
+
"global_step": 95824,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 220491 |
"eval_steps_per_second": 0.695,
|
| 220492 |
"eval_wer": 0.18689181453921008,
|
| 220493 |
"step": 95203
|
| 220494 |
+
},
|
| 220495 |
+
{
|
| 220496 |
+
"epoch": 761.02,
|
| 220497 |
+
"learning_rate": 8.490913461538462e-06,
|
| 220498 |
+
"loss": 0.3735,
|
| 220499 |
+
"step": 95205
|
| 220500 |
+
},
|
| 220501 |
+
{
|
| 220502 |
+
"epoch": 761.06,
|
| 220503 |
+
"learning_rate": 8.490833333333333e-06,
|
| 220504 |
+
"loss": 0.281,
|
| 220505 |
+
"step": 95210
|
| 220506 |
+
},
|
| 220507 |
+
{
|
| 220508 |
+
"epoch": 761.1,
|
| 220509 |
+
"learning_rate": 8.490753205128207e-06,
|
| 220510 |
+
"loss": 0.3138,
|
| 220511 |
+
"step": 95215
|
| 220512 |
+
},
|
| 220513 |
+
{
|
| 220514 |
+
"epoch": 761.14,
|
| 220515 |
+
"learning_rate": 8.490673076923078e-06,
|
| 220516 |
+
"loss": 0.3501,
|
| 220517 |
+
"step": 95220
|
| 220518 |
+
},
|
| 220519 |
+
{
|
| 220520 |
+
"epoch": 761.18,
|
| 220521 |
+
"learning_rate": 8.49059294871795e-06,
|
| 220522 |
+
"loss": 0.674,
|
| 220523 |
+
"step": 95225
|
| 220524 |
+
},
|
| 220525 |
+
{
|
| 220526 |
+
"epoch": 761.22,
|
| 220527 |
+
"learning_rate": 8.49051282051282e-06,
|
| 220528 |
+
"loss": 1.1776,
|
| 220529 |
+
"step": 95230
|
| 220530 |
+
},
|
| 220531 |
+
{
|
| 220532 |
+
"epoch": 761.26,
|
| 220533 |
+
"learning_rate": 8.490432692307694e-06,
|
| 220534 |
+
"loss": 0.2888,
|
| 220535 |
+
"step": 95235
|
| 220536 |
+
},
|
| 220537 |
+
{
|
| 220538 |
+
"epoch": 761.3,
|
| 220539 |
+
"learning_rate": 8.490352564102565e-06,
|
| 220540 |
+
"loss": 0.2533,
|
| 220541 |
+
"step": 95240
|
| 220542 |
+
},
|
| 220543 |
+
{
|
| 220544 |
+
"epoch": 761.34,
|
| 220545 |
+
"learning_rate": 8.490272435897436e-06,
|
| 220546 |
+
"loss": 0.3197,
|
| 220547 |
+
"step": 95245
|
| 220548 |
+
},
|
| 220549 |
+
{
|
| 220550 |
+
"epoch": 761.38,
|
| 220551 |
+
"learning_rate": 8.49019230769231e-06,
|
| 220552 |
+
"loss": 0.6192,
|
| 220553 |
+
"step": 95250
|
| 220554 |
+
},
|
| 220555 |
+
{
|
| 220556 |
+
"epoch": 761.42,
|
| 220557 |
+
"learning_rate": 8.490112179487179e-06,
|
| 220558 |
+
"loss": 1.0574,
|
| 220559 |
+
"step": 95255
|
| 220560 |
+
},
|
| 220561 |
+
{
|
| 220562 |
+
"epoch": 761.46,
|
| 220563 |
+
"learning_rate": 8.490032051282052e-06,
|
| 220564 |
+
"loss": 0.3482,
|
| 220565 |
+
"step": 95260
|
| 220566 |
+
},
|
| 220567 |
+
{
|
| 220568 |
+
"epoch": 761.5,
|
| 220569 |
+
"learning_rate": 8.489951923076924e-06,
|
| 220570 |
+
"loss": 0.3048,
|
| 220571 |
+
"step": 95265
|
| 220572 |
+
},
|
| 220573 |
+
{
|
| 220574 |
+
"epoch": 761.54,
|
| 220575 |
+
"learning_rate": 8.489871794871795e-06,
|
| 220576 |
+
"loss": 0.3254,
|
| 220577 |
+
"step": 95270
|
| 220578 |
+
},
|
| 220579 |
+
{
|
| 220580 |
+
"epoch": 761.58,
|
| 220581 |
+
"learning_rate": 8.489791666666666e-06,
|
| 220582 |
+
"loss": 0.7295,
|
| 220583 |
+
"step": 95275
|
| 220584 |
+
},
|
| 220585 |
+
{
|
| 220586 |
+
"epoch": 761.62,
|
| 220587 |
+
"learning_rate": 8.48971153846154e-06,
|
| 220588 |
+
"loss": 0.9835,
|
| 220589 |
+
"step": 95280
|
| 220590 |
+
},
|
| 220591 |
+
{
|
| 220592 |
+
"epoch": 761.66,
|
| 220593 |
+
"learning_rate": 8.48963141025641e-06,
|
| 220594 |
+
"loss": 0.3579,
|
| 220595 |
+
"step": 95285
|
| 220596 |
+
},
|
| 220597 |
+
{
|
| 220598 |
+
"epoch": 761.7,
|
| 220599 |
+
"learning_rate": 8.489551282051282e-06,
|
| 220600 |
+
"loss": 0.2816,
|
| 220601 |
+
"step": 95290
|
| 220602 |
+
},
|
| 220603 |
+
{
|
| 220604 |
+
"epoch": 761.74,
|
| 220605 |
+
"learning_rate": 8.489471153846155e-06,
|
| 220606 |
+
"loss": 0.345,
|
| 220607 |
+
"step": 95295
|
| 220608 |
+
},
|
| 220609 |
+
{
|
| 220610 |
+
"epoch": 761.78,
|
| 220611 |
+
"learning_rate": 8.489391025641026e-06,
|
| 220612 |
+
"loss": 0.5479,
|
| 220613 |
+
"step": 95300
|
| 220614 |
+
},
|
| 220615 |
+
{
|
| 220616 |
+
"epoch": 761.82,
|
| 220617 |
+
"learning_rate": 8.489310897435898e-06,
|
| 220618 |
+
"loss": 1.0586,
|
| 220619 |
+
"step": 95305
|
| 220620 |
+
},
|
| 220621 |
+
{
|
| 220622 |
+
"epoch": 761.86,
|
| 220623 |
+
"learning_rate": 8.489230769230769e-06,
|
| 220624 |
+
"loss": 0.3759,
|
| 220625 |
+
"step": 95310
|
| 220626 |
+
},
|
| 220627 |
+
{
|
| 220628 |
+
"epoch": 761.9,
|
| 220629 |
+
"learning_rate": 8.489150641025642e-06,
|
| 220630 |
+
"loss": 0.2978,
|
| 220631 |
+
"step": 95315
|
| 220632 |
+
},
|
| 220633 |
+
{
|
| 220634 |
+
"epoch": 761.94,
|
| 220635 |
+
"learning_rate": 8.489070512820514e-06,
|
| 220636 |
+
"loss": 0.4061,
|
| 220637 |
+
"step": 95320
|
| 220638 |
+
},
|
| 220639 |
+
{
|
| 220640 |
+
"epoch": 761.98,
|
| 220641 |
+
"learning_rate": 8.488990384615385e-06,
|
| 220642 |
+
"loss": 0.634,
|
| 220643 |
+
"step": 95325
|
| 220644 |
+
},
|
| 220645 |
+
{
|
| 220646 |
+
"epoch": 762.0,
|
| 220647 |
+
"eval_loss": 0.3908168375492096,
|
| 220648 |
+
"eval_runtime": 40.2724,
|
| 220649 |
+
"eval_samples_per_second": 20.858,
|
| 220650 |
+
"eval_steps_per_second": 0.67,
|
| 220651 |
+
"eval_wer": 0.19241309678349922,
|
| 220652 |
+
"step": 95328
|
| 220653 |
+
},
|
| 220654 |
+
{
|
| 220655 |
+
"epoch": 768.02,
|
| 220656 |
+
"learning_rate": 8.488910256410256e-06,
|
| 220657 |
+
"loss": 0.3328,
|
| 220658 |
+
"step": 95330
|
| 220659 |
+
},
|
| 220660 |
+
{
|
| 220661 |
+
"epoch": 768.06,
|
| 220662 |
+
"learning_rate": 8.48883012820513e-06,
|
| 220663 |
+
"loss": 0.307,
|
| 220664 |
+
"step": 95335
|
| 220665 |
+
},
|
| 220666 |
+
{
|
| 220667 |
+
"epoch": 768.1,
|
| 220668 |
+
"learning_rate": 8.48875e-06,
|
| 220669 |
+
"loss": 0.3208,
|
| 220670 |
+
"step": 95340
|
| 220671 |
+
},
|
| 220672 |
+
{
|
| 220673 |
+
"epoch": 768.14,
|
| 220674 |
+
"learning_rate": 8.488669871794872e-06,
|
| 220675 |
+
"loss": 0.4314,
|
| 220676 |
+
"step": 95345
|
| 220677 |
+
},
|
| 220678 |
+
{
|
| 220679 |
+
"epoch": 768.18,
|
| 220680 |
+
"learning_rate": 8.488589743589745e-06,
|
| 220681 |
+
"loss": 0.5051,
|
| 220682 |
+
"step": 95350
|
| 220683 |
+
},
|
| 220684 |
+
{
|
| 220685 |
+
"epoch": 768.22,
|
| 220686 |
+
"learning_rate": 8.488509615384616e-06,
|
| 220687 |
+
"loss": 0.9168,
|
| 220688 |
+
"step": 95355
|
| 220689 |
+
},
|
| 220690 |
+
{
|
| 220691 |
+
"epoch": 768.26,
|
| 220692 |
+
"learning_rate": 8.488429487179488e-06,
|
| 220693 |
+
"loss": 0.2612,
|
| 220694 |
+
"step": 95360
|
| 220695 |
+
},
|
| 220696 |
+
{
|
| 220697 |
+
"epoch": 768.3,
|
| 220698 |
+
"learning_rate": 8.48834935897436e-06,
|
| 220699 |
+
"loss": 0.2868,
|
| 220700 |
+
"step": 95365
|
| 220701 |
+
},
|
| 220702 |
+
{
|
| 220703 |
+
"epoch": 768.34,
|
| 220704 |
+
"learning_rate": 8.488269230769232e-06,
|
| 220705 |
+
"loss": 0.351,
|
| 220706 |
+
"step": 95370
|
| 220707 |
+
},
|
| 220708 |
+
{
|
| 220709 |
+
"epoch": 768.38,
|
| 220710 |
+
"learning_rate": 8.488189102564104e-06,
|
| 220711 |
+
"loss": 0.7034,
|
| 220712 |
+
"step": 95375
|
| 220713 |
+
},
|
| 220714 |
+
{
|
| 220715 |
+
"epoch": 768.42,
|
| 220716 |
+
"learning_rate": 8.488108974358975e-06,
|
| 220717 |
+
"loss": 1.0855,
|
| 220718 |
+
"step": 95380
|
| 220719 |
+
},
|
| 220720 |
+
{
|
| 220721 |
+
"epoch": 768.46,
|
| 220722 |
+
"learning_rate": 8.488028846153848e-06,
|
| 220723 |
+
"loss": 0.2586,
|
| 220724 |
+
"step": 95385
|
| 220725 |
+
},
|
| 220726 |
+
{
|
| 220727 |
+
"epoch": 768.5,
|
| 220728 |
+
"learning_rate": 8.48794871794872e-06,
|
| 220729 |
+
"loss": 0.2588,
|
| 220730 |
+
"step": 95390
|
| 220731 |
+
},
|
| 220732 |
+
{
|
| 220733 |
+
"epoch": 768.54,
|
| 220734 |
+
"learning_rate": 8.48786858974359e-06,
|
| 220735 |
+
"loss": 0.391,
|
| 220736 |
+
"step": 95395
|
| 220737 |
+
},
|
| 220738 |
+
{
|
| 220739 |
+
"epoch": 768.58,
|
| 220740 |
+
"learning_rate": 8.487788461538462e-06,
|
| 220741 |
+
"loss": 0.6738,
|
| 220742 |
+
"step": 95400
|
| 220743 |
+
},
|
| 220744 |
+
{
|
| 220745 |
+
"epoch": 768.62,
|
| 220746 |
+
"learning_rate": 8.487708333333335e-06,
|
| 220747 |
+
"loss": 1.0473,
|
| 220748 |
+
"step": 95405
|
| 220749 |
+
},
|
| 220750 |
+
{
|
| 220751 |
+
"epoch": 768.66,
|
| 220752 |
+
"learning_rate": 8.487628205128205e-06,
|
| 220753 |
+
"loss": 0.2548,
|
| 220754 |
+
"step": 95410
|
| 220755 |
+
},
|
| 220756 |
+
{
|
| 220757 |
+
"epoch": 768.7,
|
| 220758 |
+
"learning_rate": 8.487548076923078e-06,
|
| 220759 |
+
"loss": 0.3107,
|
| 220760 |
+
"step": 95415
|
| 220761 |
+
},
|
| 220762 |
+
{
|
| 220763 |
+
"epoch": 768.74,
|
| 220764 |
+
"learning_rate": 8.48746794871795e-06,
|
| 220765 |
+
"loss": 0.3829,
|
| 220766 |
+
"step": 95420
|
| 220767 |
+
},
|
| 220768 |
+
{
|
| 220769 |
+
"epoch": 768.78,
|
| 220770 |
+
"learning_rate": 8.48738782051282e-06,
|
| 220771 |
+
"loss": 0.6175,
|
| 220772 |
+
"step": 95425
|
| 220773 |
+
},
|
| 220774 |
+
{
|
| 220775 |
+
"epoch": 768.82,
|
| 220776 |
+
"learning_rate": 8.487307692307692e-06,
|
| 220777 |
+
"loss": 1.1168,
|
| 220778 |
+
"step": 95430
|
| 220779 |
+
},
|
| 220780 |
+
{
|
| 220781 |
+
"epoch": 768.86,
|
| 220782 |
+
"learning_rate": 8.487227564102565e-06,
|
| 220783 |
+
"loss": 0.2852,
|
| 220784 |
+
"step": 95435
|
| 220785 |
+
},
|
| 220786 |
+
{
|
| 220787 |
+
"epoch": 768.9,
|
| 220788 |
+
"learning_rate": 8.487147435897436e-06,
|
| 220789 |
+
"loss": 0.2895,
|
| 220790 |
+
"step": 95440
|
| 220791 |
+
},
|
| 220792 |
+
{
|
| 220793 |
+
"epoch": 768.94,
|
| 220794 |
+
"learning_rate": 8.487067307692308e-06,
|
| 220795 |
+
"loss": 0.3991,
|
| 220796 |
+
"step": 95445
|
| 220797 |
+
},
|
| 220798 |
+
{
|
| 220799 |
+
"epoch": 768.98,
|
| 220800 |
+
"learning_rate": 8.48698717948718e-06,
|
| 220801 |
+
"loss": 0.7499,
|
| 220802 |
+
"step": 95450
|
| 220803 |
+
},
|
| 220804 |
+
{
|
| 220805 |
+
"epoch": 769.0,
|
| 220806 |
+
"eval_loss": 0.33350133895874023,
|
| 220807 |
+
"eval_runtime": 39.8873,
|
| 220808 |
+
"eval_samples_per_second": 21.059,
|
| 220809 |
+
"eval_steps_per_second": 0.677,
|
| 220810 |
+
"eval_wer": 0.18204704020714954,
|
| 220811 |
+
"step": 95452
|
| 220812 |
+
},
|
| 220813 |
+
{
|
| 220814 |
+
"epoch": 769.02,
|
| 220815 |
+
"learning_rate": 8.486907051282052e-06,
|
| 220816 |
+
"loss": 0.3954,
|
| 220817 |
+
"step": 95455
|
| 220818 |
+
},
|
| 220819 |
+
{
|
| 220820 |
+
"epoch": 769.06,
|
| 220821 |
+
"learning_rate": 8.486826923076923e-06,
|
| 220822 |
+
"loss": 0.2815,
|
| 220823 |
+
"step": 95460
|
| 220824 |
+
},
|
| 220825 |
+
{
|
| 220826 |
+
"epoch": 769.1,
|
| 220827 |
+
"learning_rate": 8.486746794871795e-06,
|
| 220828 |
+
"loss": 0.2679,
|
| 220829 |
+
"step": 95465
|
| 220830 |
+
},
|
| 220831 |
+
{
|
| 220832 |
+
"epoch": 769.14,
|
| 220833 |
+
"learning_rate": 8.486666666666668e-06,
|
| 220834 |
+
"loss": 0.3121,
|
| 220835 |
+
"step": 95470
|
| 220836 |
+
},
|
| 220837 |
+
{
|
| 220838 |
+
"epoch": 769.18,
|
| 220839 |
+
"learning_rate": 8.48658653846154e-06,
|
| 220840 |
+
"loss": 0.6884,
|
| 220841 |
+
"step": 95475
|
| 220842 |
+
},
|
| 220843 |
+
{
|
| 220844 |
+
"epoch": 769.22,
|
| 220845 |
+
"learning_rate": 8.48650641025641e-06,
|
| 220846 |
+
"loss": 0.948,
|
| 220847 |
+
"step": 95480
|
| 220848 |
+
},
|
| 220849 |
+
{
|
| 220850 |
+
"epoch": 769.27,
|
| 220851 |
+
"learning_rate": 8.486426282051284e-06,
|
| 220852 |
+
"loss": 0.2971,
|
| 220853 |
+
"step": 95485
|
| 220854 |
+
},
|
| 220855 |
+
{
|
| 220856 |
+
"epoch": 769.31,
|
| 220857 |
+
"learning_rate": 8.486346153846155e-06,
|
| 220858 |
+
"loss": 0.369,
|
| 220859 |
+
"step": 95490
|
| 220860 |
+
},
|
| 220861 |
+
{
|
| 220862 |
+
"epoch": 769.35,
|
| 220863 |
+
"learning_rate": 8.486266025641026e-06,
|
| 220864 |
+
"loss": 0.3503,
|
| 220865 |
+
"step": 95495
|
| 220866 |
+
},
|
| 220867 |
+
{
|
| 220868 |
+
"epoch": 769.39,
|
| 220869 |
+
"learning_rate": 8.486185897435898e-06,
|
| 220870 |
+
"loss": 0.79,
|
| 220871 |
+
"step": 95500
|
| 220872 |
+
},
|
| 220873 |
+
{
|
| 220874 |
+
"epoch": 769.43,
|
| 220875 |
+
"learning_rate": 8.48610576923077e-06,
|
| 220876 |
+
"loss": 0.8587,
|
| 220877 |
+
"step": 95505
|
| 220878 |
+
},
|
| 220879 |
+
{
|
| 220880 |
+
"epoch": 769.47,
|
| 220881 |
+
"learning_rate": 8.486025641025642e-06,
|
| 220882 |
+
"loss": 0.2279,
|
| 220883 |
+
"step": 95510
|
| 220884 |
+
},
|
| 220885 |
+
{
|
| 220886 |
+
"epoch": 769.51,
|
| 220887 |
+
"learning_rate": 8.485945512820513e-06,
|
| 220888 |
+
"loss": 0.2907,
|
| 220889 |
+
"step": 95515
|
| 220890 |
+
},
|
| 220891 |
+
{
|
| 220892 |
+
"epoch": 769.55,
|
| 220893 |
+
"learning_rate": 8.485865384615385e-06,
|
| 220894 |
+
"loss": 0.34,
|
| 220895 |
+
"step": 95520
|
| 220896 |
+
},
|
| 220897 |
+
{
|
| 220898 |
+
"epoch": 769.59,
|
| 220899 |
+
"learning_rate": 8.485785256410258e-06,
|
| 220900 |
+
"loss": 0.7663,
|
| 220901 |
+
"step": 95525
|
| 220902 |
+
},
|
| 220903 |
+
{
|
| 220904 |
+
"epoch": 769.63,
|
| 220905 |
+
"learning_rate": 8.485705128205128e-06,
|
| 220906 |
+
"loss": 0.9137,
|
| 220907 |
+
"step": 95530
|
| 220908 |
+
},
|
| 220909 |
+
{
|
| 220910 |
+
"epoch": 769.67,
|
| 220911 |
+
"learning_rate": 8.485625e-06,
|
| 220912 |
+
"loss": 0.3007,
|
| 220913 |
+
"step": 95535
|
| 220914 |
+
},
|
| 220915 |
+
{
|
| 220916 |
+
"epoch": 769.71,
|
| 220917 |
+
"learning_rate": 8.485544871794874e-06,
|
| 220918 |
+
"loss": 0.3495,
|
| 220919 |
+
"step": 95540
|
| 220920 |
+
},
|
| 220921 |
+
{
|
| 220922 |
+
"epoch": 769.75,
|
| 220923 |
+
"learning_rate": 8.485464743589743e-06,
|
| 220924 |
+
"loss": 0.4196,
|
| 220925 |
+
"step": 95545
|
| 220926 |
+
},
|
| 220927 |
+
{
|
| 220928 |
+
"epoch": 769.79,
|
| 220929 |
+
"learning_rate": 8.485384615384616e-06,
|
| 220930 |
+
"loss": 0.7906,
|
| 220931 |
+
"step": 95550
|
| 220932 |
+
},
|
| 220933 |
+
{
|
| 220934 |
+
"epoch": 769.83,
|
| 220935 |
+
"learning_rate": 8.485304487179488e-06,
|
| 220936 |
+
"loss": 0.9746,
|
| 220937 |
+
"step": 95555
|
| 220938 |
+
},
|
| 220939 |
+
{
|
| 220940 |
+
"epoch": 769.87,
|
| 220941 |
+
"learning_rate": 8.485224358974359e-06,
|
| 220942 |
+
"loss": 0.2873,
|
| 220943 |
+
"step": 95560
|
| 220944 |
+
},
|
| 220945 |
+
{
|
| 220946 |
+
"epoch": 769.91,
|
| 220947 |
+
"learning_rate": 8.48514423076923e-06,
|
| 220948 |
+
"loss": 0.291,
|
| 220949 |
+
"step": 95565
|
| 220950 |
+
},
|
| 220951 |
+
{
|
| 220952 |
+
"epoch": 769.95,
|
| 220953 |
+
"learning_rate": 8.485064102564103e-06,
|
| 220954 |
+
"loss": 0.4418,
|
| 220955 |
+
"step": 95570
|
| 220956 |
+
},
|
| 220957 |
+
{
|
| 220958 |
+
"epoch": 769.99,
|
| 220959 |
+
"learning_rate": 8.484983974358975e-06,
|
| 220960 |
+
"loss": 0.7583,
|
| 220961 |
+
"step": 95575
|
| 220962 |
+
},
|
| 220963 |
+
{
|
| 220964 |
+
"epoch": 770.0,
|
| 220965 |
+
"eval_loss": 0.37961849570274353,
|
| 220966 |
+
"eval_runtime": 38.0095,
|
| 220967 |
+
"eval_samples_per_second": 22.126,
|
| 220968 |
+
"eval_steps_per_second": 0.71,
|
| 220969 |
+
"eval_wer": 0.18243194192377496,
|
| 220970 |
+
"step": 95576
|
| 220971 |
+
},
|
| 220972 |
+
{
|
| 220973 |
+
"epoch": 770.03,
|
| 220974 |
+
"learning_rate": 8.484903846153846e-06,
|
| 220975 |
+
"loss": 0.3532,
|
| 220976 |
+
"step": 95580
|
| 220977 |
+
},
|
| 220978 |
+
{
|
| 220979 |
+
"epoch": 770.07,
|
| 220980 |
+
"learning_rate": 8.48482371794872e-06,
|
| 220981 |
+
"loss": 0.2813,
|
| 220982 |
+
"step": 95585
|
| 220983 |
+
},
|
| 220984 |
+
{
|
| 220985 |
+
"epoch": 770.11,
|
| 220986 |
+
"learning_rate": 8.48474358974359e-06,
|
| 220987 |
+
"loss": 0.2872,
|
| 220988 |
+
"step": 95590
|
| 220989 |
+
},
|
| 220990 |
+
{
|
| 220991 |
+
"epoch": 770.15,
|
| 220992 |
+
"learning_rate": 8.484663461538462e-06,
|
| 220993 |
+
"loss": 0.4089,
|
| 220994 |
+
"step": 95595
|
| 220995 |
+
},
|
| 220996 |
+
{
|
| 220997 |
+
"epoch": 770.19,
|
| 220998 |
+
"learning_rate": 8.484583333333333e-06,
|
| 220999 |
+
"loss": 0.8651,
|
| 221000 |
+
"step": 95600
|
| 221001 |
+
},
|
| 221002 |
+
{
|
| 221003 |
+
"epoch": 770.23,
|
| 221004 |
+
"learning_rate": 8.484503205128206e-06,
|
| 221005 |
+
"loss": 0.834,
|
| 221006 |
+
"step": 95605
|
| 221007 |
+
},
|
| 221008 |
+
{
|
| 221009 |
+
"epoch": 770.27,
|
| 221010 |
+
"learning_rate": 8.484423076923078e-06,
|
| 221011 |
+
"loss": 0.4655,
|
| 221012 |
+
"step": 95610
|
| 221013 |
+
},
|
| 221014 |
+
{
|
| 221015 |
+
"epoch": 770.31,
|
| 221016 |
+
"learning_rate": 8.484342948717949e-06,
|
| 221017 |
+
"loss": 0.3295,
|
| 221018 |
+
"step": 95615
|
| 221019 |
+
},
|
| 221020 |
+
{
|
| 221021 |
+
"epoch": 770.35,
|
| 221022 |
+
"learning_rate": 8.48426282051282e-06,
|
| 221023 |
+
"loss": 0.3905,
|
| 221024 |
+
"step": 95620
|
| 221025 |
+
},
|
| 221026 |
+
{
|
| 221027 |
+
"epoch": 770.39,
|
| 221028 |
+
"learning_rate": 8.484182692307694e-06,
|
| 221029 |
+
"loss": 1.058,
|
| 221030 |
+
"step": 95625
|
| 221031 |
+
},
|
| 221032 |
+
{
|
| 221033 |
+
"epoch": 770.43,
|
| 221034 |
+
"learning_rate": 8.484102564102565e-06,
|
| 221035 |
+
"loss": 0.5864,
|
| 221036 |
+
"step": 95630
|
| 221037 |
+
},
|
| 221038 |
+
{
|
| 221039 |
+
"epoch": 770.47,
|
| 221040 |
+
"learning_rate": 8.484022435897436e-06,
|
| 221041 |
+
"loss": 0.2882,
|
| 221042 |
+
"step": 95635
|
| 221043 |
+
},
|
| 221044 |
+
{
|
| 221045 |
+
"epoch": 770.51,
|
| 221046 |
+
"learning_rate": 8.48394230769231e-06,
|
| 221047 |
+
"loss": 0.3196,
|
| 221048 |
+
"step": 95640
|
| 221049 |
+
},
|
| 221050 |
+
{
|
| 221051 |
+
"epoch": 770.55,
|
| 221052 |
+
"learning_rate": 8.48386217948718e-06,
|
| 221053 |
+
"loss": 0.437,
|
| 221054 |
+
"step": 95645
|
| 221055 |
+
},
|
| 221056 |
+
{
|
| 221057 |
+
"epoch": 770.59,
|
| 221058 |
+
"learning_rate": 8.483782051282052e-06,
|
| 221059 |
+
"loss": 0.9501,
|
| 221060 |
+
"step": 95650
|
| 221061 |
+
},
|
| 221062 |
+
{
|
| 221063 |
+
"epoch": 770.63,
|
| 221064 |
+
"learning_rate": 8.483701923076923e-06,
|
| 221065 |
+
"loss": 0.7306,
|
| 221066 |
+
"step": 95655
|
| 221067 |
+
},
|
| 221068 |
+
{
|
| 221069 |
+
"epoch": 770.67,
|
| 221070 |
+
"learning_rate": 8.483621794871796e-06,
|
| 221071 |
+
"loss": 0.3199,
|
| 221072 |
+
"step": 95660
|
| 221073 |
+
},
|
| 221074 |
+
{
|
| 221075 |
+
"epoch": 770.71,
|
| 221076 |
+
"learning_rate": 8.483541666666668e-06,
|
| 221077 |
+
"loss": 0.2947,
|
| 221078 |
+
"step": 95665
|
| 221079 |
+
},
|
| 221080 |
+
{
|
| 221081 |
+
"epoch": 770.76,
|
| 221082 |
+
"learning_rate": 8.483461538461539e-06,
|
| 221083 |
+
"loss": 0.3722,
|
| 221084 |
+
"step": 95670
|
| 221085 |
+
},
|
| 221086 |
+
{
|
| 221087 |
+
"epoch": 770.8,
|
| 221088 |
+
"learning_rate": 8.48338141025641e-06,
|
| 221089 |
+
"loss": 0.8973,
|
| 221090 |
+
"step": 95675
|
| 221091 |
+
},
|
| 221092 |
+
{
|
| 221093 |
+
"epoch": 770.84,
|
| 221094 |
+
"learning_rate": 8.483301282051284e-06,
|
| 221095 |
+
"loss": 0.6271,
|
| 221096 |
+
"step": 95680
|
| 221097 |
+
},
|
| 221098 |
+
{
|
| 221099 |
+
"epoch": 770.88,
|
| 221100 |
+
"learning_rate": 8.483221153846155e-06,
|
| 221101 |
+
"loss": 0.2649,
|
| 221102 |
+
"step": 95685
|
| 221103 |
+
},
|
| 221104 |
+
{
|
| 221105 |
+
"epoch": 770.92,
|
| 221106 |
+
"learning_rate": 8.483141025641026e-06,
|
| 221107 |
+
"loss": 0.3043,
|
| 221108 |
+
"step": 95690
|
| 221109 |
+
},
|
| 221110 |
+
{
|
| 221111 |
+
"epoch": 770.96,
|
| 221112 |
+
"learning_rate": 8.4830608974359e-06,
|
| 221113 |
+
"loss": 0.4054,
|
| 221114 |
+
"step": 95695
|
| 221115 |
+
},
|
| 221116 |
+
{
|
| 221117 |
+
"epoch": 771.0,
|
| 221118 |
+
"learning_rate": 8.482980769230769e-06,
|
| 221119 |
+
"loss": 1.0171,
|
| 221120 |
+
"step": 95700
|
| 221121 |
+
},
|
| 221122 |
+
{
|
| 221123 |
+
"epoch": 771.0,
|
| 221124 |
+
"eval_loss": 0.3492157757282257,
|
| 221125 |
+
"eval_runtime": 38.1524,
|
| 221126 |
+
"eval_samples_per_second": 22.017,
|
| 221127 |
+
"eval_steps_per_second": 0.708,
|
| 221128 |
+
"eval_wer": 0.18850003653101483,
|
| 221129 |
+
"step": 95700
|
| 221130 |
+
},
|
| 221131 |
+
{
|
| 221132 |
+
"epoch": 771.04,
|
| 221133 |
+
"learning_rate": 8.482900641025642e-06,
|
| 221134 |
+
"loss": 0.361,
|
| 221135 |
+
"step": 95705
|
| 221136 |
+
},
|
| 221137 |
+
{
|
| 221138 |
+
"epoch": 771.08,
|
| 221139 |
+
"learning_rate": 8.482820512820513e-06,
|
| 221140 |
+
"loss": 0.2913,
|
| 221141 |
+
"step": 95710
|
| 221142 |
+
},
|
| 221143 |
+
{
|
| 221144 |
+
"epoch": 771.12,
|
| 221145 |
+
"learning_rate": 8.482740384615385e-06,
|
| 221146 |
+
"loss": 0.3224,
|
| 221147 |
+
"step": 95715
|
| 221148 |
+
},
|
| 221149 |
+
{
|
| 221150 |
+
"epoch": 771.16,
|
| 221151 |
+
"learning_rate": 8.482660256410256e-06,
|
| 221152 |
+
"loss": 0.4519,
|
| 221153 |
+
"step": 95720
|
| 221154 |
+
},
|
| 221155 |
+
{
|
| 221156 |
+
"epoch": 771.2,
|
| 221157 |
+
"learning_rate": 8.48258012820513e-06,
|
| 221158 |
+
"loss": 1.2417,
|
| 221159 |
+
"step": 95725
|
| 221160 |
+
},
|
| 221161 |
+
{
|
| 221162 |
+
"epoch": 771.24,
|
| 221163 |
+
"learning_rate": 8.4825e-06,
|
| 221164 |
+
"loss": 0.3535,
|
| 221165 |
+
"step": 95730
|
| 221166 |
+
},
|
| 221167 |
+
{
|
| 221168 |
+
"epoch": 771.28,
|
| 221169 |
+
"learning_rate": 8.482419871794872e-06,
|
| 221170 |
+
"loss": 0.321,
|
| 221171 |
+
"step": 95735
|
| 221172 |
+
},
|
| 221173 |
+
{
|
| 221174 |
+
"epoch": 771.32,
|
| 221175 |
+
"learning_rate": 8.482339743589745e-06,
|
| 221176 |
+
"loss": 0.3113,
|
| 221177 |
+
"step": 95740
|
| 221178 |
+
},
|
| 221179 |
+
{
|
| 221180 |
+
"epoch": 771.36,
|
| 221181 |
+
"learning_rate": 8.482259615384616e-06,
|
| 221182 |
+
"loss": 0.4148,
|
| 221183 |
+
"step": 95745
|
| 221184 |
+
},
|
| 221185 |
+
{
|
| 221186 |
+
"epoch": 771.4,
|
| 221187 |
+
"learning_rate": 8.482179487179488e-06,
|
| 221188 |
+
"loss": 1.2129,
|
| 221189 |
+
"step": 95750
|
| 221190 |
+
},
|
| 221191 |
+
{
|
| 221192 |
+
"epoch": 771.44,
|
| 221193 |
+
"learning_rate": 8.482099358974359e-06,
|
| 221194 |
+
"loss": 0.3261,
|
| 221195 |
+
"step": 95755
|
| 221196 |
+
},
|
| 221197 |
+
{
|
| 221198 |
+
"epoch": 771.48,
|
| 221199 |
+
"learning_rate": 8.482019230769232e-06,
|
| 221200 |
+
"loss": 0.274,
|
| 221201 |
+
"step": 95760
|
| 221202 |
+
},
|
| 221203 |
+
{
|
| 221204 |
+
"epoch": 771.52,
|
| 221205 |
+
"learning_rate": 8.481939102564103e-06,
|
| 221206 |
+
"loss": 0.2832,
|
| 221207 |
+
"step": 95765
|
| 221208 |
+
},
|
| 221209 |
+
{
|
| 221210 |
+
"epoch": 771.56,
|
| 221211 |
+
"learning_rate": 8.481858974358975e-06,
|
| 221212 |
+
"loss": 0.4462,
|
| 221213 |
+
"step": 95770
|
| 221214 |
+
},
|
| 221215 |
+
{
|
| 221216 |
+
"epoch": 771.6,
|
| 221217 |
+
"learning_rate": 8.481778846153846e-06,
|
| 221218 |
+
"loss": 1.3003,
|
| 221219 |
+
"step": 95775
|
| 221220 |
+
},
|
| 221221 |
+
{
|
| 221222 |
+
"epoch": 771.64,
|
| 221223 |
+
"learning_rate": 8.48169871794872e-06,
|
| 221224 |
+
"loss": 0.3361,
|
| 221225 |
+
"step": 95780
|
| 221226 |
+
},
|
| 221227 |
+
{
|
| 221228 |
+
"epoch": 771.68,
|
| 221229 |
+
"learning_rate": 8.48161858974359e-06,
|
| 221230 |
+
"loss": 0.271,
|
| 221231 |
+
"step": 95785
|
| 221232 |
+
},
|
| 221233 |
+
{
|
| 221234 |
+
"epoch": 771.72,
|
| 221235 |
+
"learning_rate": 8.481538461538462e-06,
|
| 221236 |
+
"loss": 0.3512,
|
| 221237 |
+
"step": 95790
|
| 221238 |
+
},
|
| 221239 |
+
{
|
| 221240 |
+
"epoch": 771.76,
|
| 221241 |
+
"learning_rate": 8.481458333333335e-06,
|
| 221242 |
+
"loss": 0.4285,
|
| 221243 |
+
"step": 95795
|
| 221244 |
+
},
|
| 221245 |
+
{
|
| 221246 |
+
"epoch": 771.8,
|
| 221247 |
+
"learning_rate": 8.481378205128206e-06,
|
| 221248 |
+
"loss": 1.1133,
|
| 221249 |
+
"step": 95800
|
| 221250 |
+
},
|
| 221251 |
+
{
|
| 221252 |
+
"epoch": 771.84,
|
| 221253 |
+
"learning_rate": 8.481298076923078e-06,
|
| 221254 |
+
"loss": 0.2918,
|
| 221255 |
+
"step": 95805
|
| 221256 |
+
},
|
| 221257 |
+
{
|
| 221258 |
+
"epoch": 771.88,
|
| 221259 |
+
"learning_rate": 8.481217948717949e-06,
|
| 221260 |
+
"loss": 0.2749,
|
| 221261 |
+
"step": 95810
|
| 221262 |
+
},
|
| 221263 |
+
{
|
| 221264 |
+
"epoch": 771.92,
|
| 221265 |
+
"learning_rate": 8.481137820512822e-06,
|
| 221266 |
+
"loss": 0.3116,
|
| 221267 |
+
"step": 95815
|
| 221268 |
+
},
|
| 221269 |
+
{
|
| 221270 |
+
"epoch": 771.96,
|
| 221271 |
+
"learning_rate": 8.481057692307693e-06,
|
| 221272 |
+
"loss": 0.5563,
|
| 221273 |
+
"step": 95820
|
| 221274 |
+
},
|
| 221275 |
+
{
|
| 221276 |
+
"epoch": 772.0,
|
| 221277 |
+
"eval_loss": 0.4196493923664093,
|
| 221278 |
+
"eval_runtime": 38.6955,
|
| 221279 |
+
"eval_samples_per_second": 21.708,
|
| 221280 |
+
"eval_steps_per_second": 0.698,
|
| 221281 |
+
"eval_wer": 0.1937148929710035,
|
| 221282 |
+
"step": 95824
|
| 221283 |
}
|
| 221284 |
],
|
| 221285 |
"max_steps": 620000,
|
| 221286 |
"num_train_epochs": 5000,
|
| 221287 |
+
"total_flos": 2.6965513860738492e+20,
|
| 221288 |
"trial_name": null,
|
| 221289 |
"trial_params": null
|
| 221290 |
}
|
model-bin/finetune/base/{checkpoint-95203 → checkpoint-95824}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629964487.380869/events.out.tfevents.1629964487.8e89bd551565.924.111
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:370fb1e5a21e5ed84af42256a50fc1559af0569d214371fd6a007326bca845a2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629964926.187713/events.out.tfevents.1629964926.8e89bd551565.924.113
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0365e2dbc6178de478f37251cd475fe7c03da2ee0cec7bd1f97fa17f347101db
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629965370.4697628/events.out.tfevents.1629965370.8e89bd551565.924.115
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:29adf051b9afb640abe92caf8706a078a74468221337af4423ef7c5c82942761
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629965797.9083533/events.out.tfevents.1629965797.8e89bd551565.924.117
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c38a0d0648d9afa30bd6a184295a5da6120c73907057de82af76d662f892710e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629966225.1204813/events.out.tfevents.1629966225.8e89bd551565.924.119
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:feb7fc5ecfbaca7f56f28251f593c5306e8cf68c422b21569b004ef7fe7db5d1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629964487.8e89bd551565.924.110
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ce0236705ca4db610d29576f7ded67f7eb44fe7a2163e4ba0315572861a2c512
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629964926.8e89bd551565.924.112
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c4230c306ec3cfc64cc1a2c0d801a88a1af60ecac44b4110d9883f70c88c60fc
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629965368.8e89bd551565.924.114
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:55863a509b5da2b4920c1bedd9fb7b552136f69591fa01f51c3ae538e771db1d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629965797.8e89bd551565.924.116
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:780534551afb252eb6467553112322bda2937c2a339d7e10befc9783e0a65572
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629966225.8e89bd551565.924.118
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f1639a881fe5098533b41b3e65cccbe304ed455d6471a834b91c59f17348cad7
|
| 3 |
+
size 8462
|